Re: [PATCH 2/9] target/ppc: add errp to kvmppc_read_int_cpu_dt()

[Mark Cave-Ayland]

On 12/07/2022 15:54, BALATON Zoltan wrote:

> On Tue, 12 Jul 2022, Mark Cave-Ayland wrote:
> > On 11/07/2022 08:42, Cédric Le Goater wrote:
> > > > > Anything special I should know ?
> > > >
> > > > As I don't have access to a G5 I've never tried that, however the 
> > > > qemu-system-ppc64 mac99 is wired differently to the qemu-system-ppc mac99 machine 
> > > > so I wouldn't be surprised if something is broken there.
> > > >
> > > > My normal test for MacOS is something like:
> > > >
> > > >     qemu-system-ppc -M mac99 -accel kvm -hda macos104.img
> > > >
> > > > Can you try qemu-system-ppc and see if it is any better? If not then I can fire 
> > > > up the G4 and get the git hashes for my last known working configuration.
> > >
> > > Same issue with 32bit.
> >
> > I've just fired up my G4 to test this again, pulled the latest QEMU git master and 
> > confirmed that I have a working setup with the details below:
> >
> > Host kernel: (5.1.0-rc2+)
> > commit a3ac7917b73070010c05b4485b8582a6c9cd69b6
> > Author: Linus Torvalds <torvalds@linux-foundation.org>
> > Date:   Mon Mar 25 14:49:00 2019 -0700
> >
> > Guest kernel: (4.14.0-3-powerpc)
> > using Debian ports debian-9.0-powerpc-NETINST-1.iso
> >
> > Command line:
> > ./qemu-system-ppc [-M mac99] -accel kvm -cdrom 
> > /home/mca/images/debian-9.0-powerpc-NETINST-1.iso -boot d -nographic
> >
> > However if I switch to using the latest Debian ports 
> > debian-10.0.0-powerpc-NETINST-1.iso then I get a failure:
> >
> > [    0.198565] BUG: Unable to handle kernel data access on read at 0xbb0030d4
>
> What is or should be at this address and why does the kernel access it? By default I 
> see nothing mapped there. Do you need more RAM? Maybe the default 128 MB is not 
> enough for newer kernels? I've seen such problem with other OSes before.

Yeah I've already tried increasing the RAM and it makes no difference. It wouldn't 
surprise me if it's a kernel issue since Christophe has done a lot of low-level work 
for 32-bit PPC including moving routines from asm to C and KASAN. I'm away for a few 
days but I will do a bisect when I get back, unless anyone beats me to it...


ATB,

Mark.